Real-Time RF Directional Tracking with Multi-Modal Fusion Benjamin J Gilbert College of the Mainland Robotic Process AutomationDownload
A lightweight, reproducible pipeline for
real-time RF directional tracking using opportunistic Wi-Fi
CSI, BLE RSSI, and UWB measurements. A six-state Kalman
filter with adaptive measurement noise fuses multi-rate observations and optionally incorporates a learned dynamics prior
(DOMA). The system exposes a FastAPI WebSocket interface
with QuestDB storage, achieving sub-30ms latency for streaming
predictions. We demonstrate 15-25% ADE improvements over
single-modality baselines and provide a complete reproducible
build with JSONTeX auto-generated tables and figures.